CN107180434A - Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic - Google Patents
Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic Download PDFInfo
- Publication number
- CN107180434A CN107180434A CN201710411372.7A CN201710411372A CN107180434A CN 107180434 A CN107180434 A CN 107180434A CN 201710411372 A CN201710411372 A CN 201710411372A CN 107180434 A CN107180434 A CN 107180434A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- pixel
- msup
- super
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/143—Segmentation; Edge detection involving probabilistic approaches, e.g. Markov random field [MRF] modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
Abstract
The invention discloses the polarimetric synthetic aperture radar based on super-pixel and fractal net work evolution algorithmic (SAR) image partition method, for Polarimetric SAR Image to be split, generation super-pixel is used as initial object；Calculate the similarity criterion between adjacent object in initial object；The minimum value of similarity criterion between each object and adjacent object is counted, if minimum value is less than or equal to scale parameter, merges two adjacent objects and generates new object, if minimum value is more than scale parameter, nonjoinder travels through all objects, completion is once split, and generates new object layer；Object polygon is generated, final segmentation result is obtained.The present invention provides enough pixels for the estimation of statistical model parameter, it is to avoid the zigzag phenomenon of partitioning boundary；Thought based on fractal net work evolution algorithmic combines statistical nature and shape facility, makes cutting object more consistent, and border is smooth, improves Polarimetric SAR Image segmentation accuracy.
Description
Technical field
The present invention relates to polarimetric synthetic aperture radar (Synthetic Aperture Radar, SAR) image partition method,
More particularly to the polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic.
Background technology
Image segmentation is exactly to divide the image into several regions specific, with unique properties and propose interesting target
Technology and process.It is by the committed step of image procossing to graphical analysis.Polarimetric SAR Image is due to by coherent speckle noise
Interference, atural object obscurity boundary, the interpretation of Polarimetric SAR Image is relatively difficult.And object-based method can not only effectively suppress phase
Dry spot influences, and also introduces more available features, helps to understand the atural object and target information that image is included, interpretation knot
Fruit is more preferably.
For the above feature of Polarimetric SAR Image, combination features segmentation is to solve one of effective ways of the problem, mesh
The preceding main utilization for considering statistical nature and spatial information of related research.
The dividing method of utilization space information is mainly comprising the Polarimetric SAR Image segmentation based on markov random file and base
Split in the Polarimetric SAR Image of fractal net work evolution algorithmic.Markov random file can preferably describe neighborhood information, but should
Class method only takes into account the spatial coherence between pixel, and the spatial form feature enriched in high-definition picture is not utilized.
FNEA can utilize the spatial form feature of target well, some scholars respectively by shape facility and H/ α/A characteristics of decomposition,
Freeman characteristics of decomposition, Pauli characteristics of decomposition, multiple polarization characteristics for building target are integrated and split.This kind of method
Good segmentation effect is achieved, but mainly uses goal decomposition feature and power to be split, without considering high-resolution
Rate full-polarization SAR covariance matrix or the distinctive statistical property of coherence matrix are split.
Mainly coherence matrix is modeled using statistical model using the dividing method of statistical nature, pushed away based on statistical model
Statistical nature similarity criterion between derived object is used for the segmentation of Polarimetric SAR Image.Different heterogeneous degree in Polarimetric SAR Image
Target there are different statistical properties.The coherence matrix in homogeneous area is generally distributed modeling, but Wishart points using Wishart
The textured heterogeneous area of the unsuitable description of cloth, the raising of particularly SAR spatial resolutions make it that atural object texture information is richer
Richness, is typically based on the product model comprising texture variable to model to heterogeneous area.Gamma is obeyed when texture variable to be distributed,
Inverse Gamma are distributed or Fisher distributions, and corresponding coherence matrix then obeys K distributions, G respectively0Distribution or KummerU divide
Cloth.This kind of method needs enough samples to estimate the parameter of product model, general to be used as initial segmentation using grid partition.Net
On the one hand lattice partitioning can cause final segmentation result zigzag substantially, on the other hand, the square of mesh generation and the side of atural object
Boundary is variant, can reduce the accuracy of parameter estimation.
The content of the invention
In view of this, The embodiment provides a kind of Polarimetric SAR Image segmentation accuracy is high based on super-pixel
With the polarization SAR image segmentation method of fractal net work evolution algorithmic.
Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic, is comprised the following steps：
Step 1：For Polarimetric SAR Image to be split, generation super-pixel is used as initial object；
Step 2：Calculate each similarity criterion between object and adjacent object in initial object；
Step 3：The minimum value of similarity criterion between each object and adjacent object is counted, if minimum value is less than or equal to
Scale parameter, then merge two adjacent objects and generate new object, if minimum value is more than scale parameter, nonjoinder, traversal is all
Object, completion is once split, and generates new object layer；
Step 4：Repeat step 2 and step 3 no longer change until object number, obtain final object layer；
Step 5：Object polygon is generated according to step 4, that is, obtains final segmentation result.
Further, in the step 1, initial object is used as based on Polarimetric SAR Image generation super-pixel.
Further, in the step 1, Pauli is carried out to Polarimetric SAR Image and decomposes generation PauliRGB images；It is comprehensive
The polarization characteristic distance and space length of PauliRGB images complete K averages as final distance metric in subrange
Iteration is clustered, and generation super-pixel is used as initial object.
Further, in the step 1, concretely comprising the following steps for super-pixel is generated with simple linear Iterative Clustering：
Step 1.1：Determine desired super-pixel size g2, and then step-length g is determined, according to step-length sampling selected seed point,
And seed point is adjusted in 3*3 subranges to the gradient smallest point of PauliRGB images；
Step 1.2：Calculated respectively in each central seed point 2g*2g subranges between each pixel and correspondence seed point
Distance；
Step 1.3：Distance between each pixel and correspondence seed point that are obtained according to step 1.2 carries out part K averages and changed
Generation cluster, until restraining or reaching maximum iteration, that is, generates super-pixel；
Step 1.4：The super-pixel that number of pixels is less than setting numerical value is merged into adjacent and closest super-pixel,
Obtain the super-pixel as initial object.
Further, in the step 1.1, the PauliRGB values of m-th seed point and locus are labeled as
Cm=[Rm Gm Bm xm ym]T；
In the step 1.2, the calculation formula of distance is between each pixel and correspondence seed point：
In formula：D represents the distance between pixel and correspondence seed point, dpThe polarization characteristic distance decomposed based on Pauli is represented,
dsFor space length, max (dp) for the maximum of the polarization characteristic distances decomposed of Pauli in the cluster, g for space in cluster away from
From maximum, with max (dp) and g respectively to polarization characteristic apart from dpWith space length dsIt is standardized.
Further, in the step 2, the similarity criterion between each object and adjacent object is drilled based on fractal net work
Change algorithm calculating to obtain, the similarity criterion between each object and adjacent object combines statistical nature and shape facility.
Further, in the step 2, each statistics similarity between object and adjacent object in initial object is calculated,
Build comprehensive statistics feature and the similarity criterion of shape facility between adjacent object.
Further, in the step 2, comprehensive statistics feature and the similarity criterion of shape facility between adjacent object are built
Specific method is：
Step 2.1：Using statistical model to polarization SAR data modeling, and estimated according to the probability density function of statistical model
Calculate the form parameter of statistical model；
Step 2.2：The heterogeneous degree of statistical property of object is calculated using likelihood function；
Step 2.3：Calculate the statistics similarity criterion between adjacent object；
Step 2.4：Will degree of compacting and smoothness setting weight, the heterogeneous degree of shape facility of object is calculated, and according to object
The heterogeneous degree of shape facility change description adjacent object between shape similarity criterion；
Step 2.5：In the multidimensional feature space that statistical nature and shape facility are constituted, the statistics that step 2.3 is obtained
The shape similarity criterion that similarity criterion and step 2.4 are obtained is weighted, and then the synthesis calculated between adjacent object is similar
Property criterion, comprehensive statistics feature and the similarity criterion of shape facility as between adjacent object.
Further, in the step 2.1, statistical model is G0Distributed model, the probability density function of statistical model is：
In formula：Σ is the desired value of coherence matrix, its estimateIt can be estimated by the average value of sample：
∑=E [T], L is that, regarding number, α is form parameter, and Γ () is Γ-function, tr () and | | represent to ask respectively the mark of matrix with
Determinant, d is coherence matrix T dimension, the d=3 under the conditions of reciprocal theorem is met；
According to the probability density function of statistical model, and based on tr (Σ-1T second order moment characteristics estimation statistical model probability)
The calculation formula of form parameter in density function is：
In formula:M=tr (∑s-1T)。
Further, in the step 2.2, the heterogeneous degree calculation formula of statistical property of object is：
In formula：N is the number of pixels of object.
In the step 2.3, the calculation formula of the statistics similarity criterion between adjacent object is：
In formula：I and j represent object i and the object j adjacent with object i respectively.
In the step 2.4, the calculation formula of the heterogeneous degree of object shapes feature is：
In formula：WcompctFor the weight of object degree of compacting, c is object bounds girth, and b is all for the minimum outsourcing rectangle of object
It is long；
It is public according to the calculating of the shape similarity criterion between the change description adjacent object of the heterogeneous degree of the shape facility of object
Formula is：
In the step 2.5, the calculation formula of the synthesis similarity criterion between adjacent object is：
Δ h=wshapeΔhshape+(1-wshape)Δhstt。
Compared with prior art, the invention has the advantages that：With the polarization characteristic based on PauliRGB apart from generation
For the color distance in the CIELAB spaces in simple linear iterative algorithm, the generation super-pixel conduct of simple linear iterative algorithm is improved
Initial object both provided enough pixels for the estimation of statistical model parameter, and it also avoid window as initial segmentation object
Partitioning boundary zigzag phenomenon caused by sample statistics parameter Estimation；Based on the thought of fractal net work evolution algorithmic, system is combined
Count feature and shape facility to calculate the similarity criterion between object, make cutting object more consistent, border is smooth, improve
The accuracy of Polarimetric SAR Image segmentation.
Brief description of the drawings
Fig. 1 is a flow of the polarization SAR image segmentation method of the invention based on super-pixel and fractal net work evolution algorithmic
Figure.
Fig. 2 is polarization SAR image segmentation method of the embodiment of the present invention 1 based on super-pixel and fractal net work evolution algorithmic
One flow chart.
Fig. 3 is polarization SAR image segmentation method of the embodiment of the present invention 2 based on super-pixel and fractal net work evolution algorithmic
One flow chart.
Fig. 4 is polarization SAR image segmentation method of the embodiment of the present invention 3 based on super-pixel and fractal net work evolution algorithmic
One flow chart.
Fig. 5 is the segmentation result for the regional Polarimetric SAR Images of Dutch Flevoland that present invention emulation is used.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing to embodiment party of the present invention
Formula is further described.
Embodiment 1
Fig. 1 and Fig. 2 are refer to, The embodiment provides the polarization based on super-pixel and fractal net work evolution algorithmic
SAR image segmentation method, is comprised the following steps：
Step 1：Pauli, which is carried out, for Polarimetric SAR Image to be split decomposes generation PauliRGB images, and with improved
Simple linear Iterative Clustering generates initial object；
It is a kind of Coherent decomposition that Pauli, which is decomposed, under reciprocity condition, the collision matrix of certainty target is decomposed into flat
Two faces of the corner reflector that scattering,single, the deflection on surface are the dihedral angle scattering of 0 ° of corner reflector and deflection is 45 °
The linear combination of angle scattering.
During statistical nature similitude due to calculating object and adjacent object, it is desirable to have enough sample estimation statistical models
Parameter, therefore the use for the statistics similarity criterion set up based on statistical model should set up and having the initial of certain number of pixels
On cutting object, the CIELAB spaces in simple linear Iterative Clustering are replaced with the polarization characteristic distance based on PauliRGB
Color distance, the polarization characteristic distance and space length of comprehensive PauliRGB images are as final distance metric, in part
In the range of complete K mean iteratives cluster, generation super-pixel is used as initial object；
Concretely comprise the following steps：
Step 1.1：Determine desired super-pixel size g2, and then step-length g is determined, according to step-length sampling selected seed point,
Impacted in order to avoid seed point falls at the edge of atural object to segmentation below, seed point is adjusted in 3*3 subranges and is arrived
The gradient smallest point of PauliRGB images；The PauliRGB values of m-th seed point and locus are labeled as Cm=[Rm Gm Bm
xm ym]T；
Step 1.2：Calculated respectively in each central seed point 2g*2g subranges between each pixel and correspondence seed point
Distance；
The calculation formula of distance is between each pixel and correspondence seed point：
In formula：D represents the distance between pixel and correspondence seed point, dpThe polarization characteristic distance decomposed based on Pauli is represented,
dsFor space length, max (dp) for the maximum of the polarization characteristic distances decomposed of Pauli in the cluster, g for space in cluster away from
From maximum, with max (dp) and g respectively to polarization characteristic apart from dpWith space length dsIt is standardized；
Step 1.3：Distance between each pixel and correspondence seed point that are obtained according to step 1.2 carries out part K averages and changed
Generation cluster, until restraining or reaching maximum iteration, that is, generates super-pixel；
Step 1.4：Because the too small parameter estimation that can make statistical model of number of pixels is inaccurate, statistics similarity can be influenceed
The accuracy of criterion, the super-pixel that number of pixels is less than setting numerical value is merged into adjacent and closest super-pixel, i.e.,
Obtain the super-pixel as initial object；
Step 2：Calculate each similarity criterion between object and adjacent object in initial object；
Each the similarity criterion between object and adjacent object is to calculate to obtain based on fractal net work evolution algorithmic, each
Similarity criterion between object and adjacent object combines statistical nature and shape facility.
Step 3：The minimum value of similarity criterion between each object and adjacent object is counted, if minimum value is less than or equal to
Scale parameter, then merge two adjacent objects and generate new object, if minimum value is more than scale parameter, nonjoinder, traversal is all
Object, completion is once split, and generates new object layer；
Step 4：Repeat step 2 and step 3 no longer change until object number, obtain final object layer；
Step 5：According to the generation object polygon of step 4, that is, obtain final segmentation result.
The color in the CIELAB spaces in simple linear iterative algorithm is replaced with the polarization characteristic distance based on PauliRGB
Distance, improves simple linear iterative algorithm and generates super-pixel as initial object as initial segmentation object, be both statistical model
The estimation of parameter provides enough pixels, it also avoid the zigzag phenomenon of partitioning boundary；Based on fractal net work evolution algorithmic
Thought, combine statistical nature and shape facility to calculate the similarity criterion between object, make cutting object more consistent,
Border is smooth, improves the accuracy of Polarimetric SAR Image segmentation.
Embodiment 2
Fig. 1 and Fig. 3 are refer to, The embodiment provides the polarization based on super-pixel and fractal net work evolution algorithmic
SAR image segmentation method, is comprised the following steps：
Step 1：For Polarimetric SAR Image to be split, generation super-pixel is used as initial object；
Step 2：The spectrum phase in fractal net work evolution algorithmic is replaced with similitude between the object derived based on statistical model
Like property, the similarity criterion for combining shape facility and statistical nature is set up, each object and phase adjacency pair in initial object is calculated
Statistics similarity as between, builds comprehensive statistics feature and the similarity criterion of shape facility between adjacent object；
Specific method is：
Step 2.1：Using statistical model to polarization SAR data modeling, preferably G0Distributed model, and according to statistical model
Probability density function estimates form parameter；
The probability density function of statistical model is：
In formula：∑ is the desired value of coherence matrix, its estimateIt can be estimated by the average value of sample：∑=E [T], L is that, regarding number, α is form parameter, and Γ () is Γ-function, tr () and | | expression is asked respectively
The mark and determinant of matrix, d are coherence matrix T dimension, the d=3 under the conditions of reciprocal theorem is met.
According to the probability density function of statistical model, and based on tr (∑s-1T second order moment characteristics estimation statistical model probability)
The calculation formula of form parameter in density function is：
In formula：M=tr (∑s-1T)
Step 2.2：The heterogeneous degree of statistical property of object is calculated using likelihood function；
The heterogeneous degree calculation formula of the statistical property of object is：
In formula：N is the number of pixels of object.
Step 2.3：Calculate the statistics similarity criterion between adjacent object；Due in FNEA cutting procedures, two are merged every time
Individual adjacent object, which produces new object, all can decline log-likelihood function value, so log-likelihood function value declines before merging
Minimum two neighboring object is merged, and the calculation formula of the statistics similarity criterion between adjacent object is：
In formula：I and j represent object i and the object j adjacent with object i respectively.
Step 2.4：Shape similarity criterion uses identical in FNEA algorithms to define, from two kinds of scapes of degree of compacting and smoothness
See ecology to estimate to define, degree of compacting characterizes the compactness of object, can be described as object bounds girth c and pixel count in object
Ratio between n root mean square；Smoothness characterizes the smooth degree of object bounds, with object bounds girth c and minimum outsourcing square
Ratio between shape girth b is described, and degree of compacting and smoothness are set into weight, calculates the heterogeneous degree of shape facility of object, and root
According to the shape similarity criterion between the change description adjacent object of the heterogeneous degree of the shape facility of object；
The calculation formula of the heterogeneous degree of object shapes feature is：
In formula：WcompctFor the weight of object degree of compacting, c is object bounds girth, and b is all for the minimum outsourcing rectangle of object
It is long.
It is public according to the calculating of the shape similarity criterion between the change description adjacent object of the heterogeneous degree of the shape facility of object
Formula is：
Step 2.5：In the multidimensional feature space that statistical nature and shape facility are constituted, the statistics that step 2.3 is obtained
The shape similarity criterion that similarity criterion and step 2.4 are obtained is weighted, and then the synthesis calculated between adjacent object is similar
Property criterion, comprehensive statistics feature and the similarity criterion of shape facility as between adjacent object.
The calculation formula of synthesis similarity criterion between adjacent object is：
Δ h=wshapeΔhshape+(1-wshape)Δhstt。
Step 3：The minimum value of similarity criterion between each object and adjacent object is counted, if minimum value is less than or equal to
Scale parameter, then merge adjacent object and generate new object, if minimum value is more than scale parameter, nonjoinder, traversal is all right
As completion is once split, and generates new object layer；
Step 4：Repeat step 2 and step 3 no longer change until object number, obtain final object layer；
Step 5：According to the generation object polygon of step 4, that is, obtain final segmentation result.
The color in the CIELAB spaces in simple linear iterative algorithm is replaced with the polarization characteristic distance based on PauliRGB
Distance, improves simple linear iterative algorithm and generates super-pixel as initial object as initial segmentation object, be both statistical model
The estimation of parameter provides enough pixels, it also avoid the zigzag phenomenon of partitioning boundary；Based on fractal net work evolution algorithmic
Thought, combine statistical nature and shape facility to calculate the similarity criterion between object, make cutting object more consistent,
Border is smooth, improves the accuracy of Polarimetric SAR Image segmentation.
Embodiment 3
Fig. 1 and Fig. 4 are refer to, The embodiment provides the polarization based on super-pixel and fractal net work evolution algorithmic
SAR image segmentation method, is comprised the following steps：
Step 1：Pauli, which is carried out, for Polarimetric SAR Image to be split decomposes generation PauliRGB images, and with improved
Simple linear Iterative Clustering generates initial object；
It is a kind of Coherent decomposition that Pauli, which is decomposed, under reciprocity condition, the collision matrix of certainty target is decomposed into flat
Two faces of the corner reflector that scattering,single, the deflection on surface are the dihedral angle scattering of 0 ° of corner reflector and deflection is 45 °
The linear combination of angle scattering.
During statistical nature similitude due to calculating object and adjacent object, it is desirable to have enough sample estimation statistical models
Parameter, therefore the use for the statistics similarity criterion set up based on statistical model should set up and having the initial of certain number of pixels
On cutting object, the CIELAB spaces in simple linear Iterative Clustering are replaced with the polarization characteristic distance based on PauliRGB
Color distance, the polarization characteristic distance and space length of comprehensive PauliRGB images are as final distance metric, in part
In the range of complete K mean iteratives cluster, generation super-pixel is used as initial object；
Concretely comprise the following steps：
Step 1.1：Determine desired super-pixel size g2, and then step-length g is determined, according to step-length sampling selected seed point,
Impacted in order to avoid seed point falls at the edge of atural object to segmentation below, seed point is adjusted in 3*3 subranges and is arrived
The gradient smallest point of PauliRGB images；The PauliRGB values of m-th seed point and locus are labeled as Cm=[Rm Gm Bm
xm ym]T；
Step 1.2：Calculated respectively in each central seed point 2g*2g subranges between each pixel and correspondence seed point
Distance；
The calculation formula of distance is between each pixel and correspondence seed point：
In formula：D represents the distance between pixel and correspondence seed point, dpThe polarization characteristic distance decomposed based on Pauli is represented,
dsFor space length, max (dp) for the maximum of the polarization characteristic distances decomposed of Pauli in the cluster, g for space in cluster away from
From maximum, with max (dp) and g respectively to polarization characteristic apart from dpWith space length dsIt is standardized；
Step 1.3：Distance between each pixel and correspondence seed point that are obtained according to step 1.2 carries out part K averages and changed
Generation cluster, until restraining or reaching maximum iteration, that is, generates super-pixel；
Step 1.4：Because the too small parameter estimation that can make statistical model of number of pixels is inaccurate, statistics similarity can be influenceed
The accuracy of criterion, the super-pixel that number of pixels is less than setting numerical value is merged into adjacent and closest super-pixel, i.e.,
Obtain the super-pixel as initial object；
Step 2：Calculate each similarity criterion between object and adjacent object in initial object；With based on statistical model
Similitude replaces the spectral similarity in fractal net work evolution algorithmic between the object of derivation, and foundation combines shape facility and statistics
The similarity criterion of feature, calculates each statistics similarity between object and adjacent object in initial object, builds phase adjacency pair
As a comprehensive statistics feature and the similarity criterion of shape facility；
Specific method is：
Step 2.1：Using statistical model to polarization SAR data modeling, preferably G0Distributed model, and according to statistical model
Probability density function estimates form parameter；
The probability density function of statistical model is：
In formula：∑ is the desired value of coherence matrix, its estimateIt can be estimated by the average value of sample：∑=E [T], L is that, regarding number, α is form parameter, and Γ () is Γ-function, tr () and | | expression is asked respectively
The mark and determinant of matrix, d are coherence matrix T dimension, the d=3 under the conditions of reciprocal theorem is met.
According to the probability density function of statistical model, and based on tr (∑s-1T second order moment characteristics estimation statistical model probability)
The calculation formula of form parameter in density function is：
In formula：M=tr (∑s-1T)。
Step 2.2：The heterogeneous degree of statistical property of object is calculated using likelihood function；
The heterogeneous degree calculation formula of the statistical property of object is：
In formula：N is the number of pixels of object.
Step 2.3：Calculate the statistics similarity criterion between adjacent object；Due in FNEA cutting procedures, two are merged every time
Individual adjacent object, which produces new object, all can decline log-likelihood function value, so log-likelihood function value declines before merging
Minimum two neighboring object is merged, and the calculation formula of the statistics similarity criterion between adjacent object is：
In formula：I and j represent object i and the object j adjacent with object i respectively.
Step 2.4：Shape similarity criterion uses identical in FNEA algorithms to define, from two kinds of scapes of degree of compacting and smoothness
See ecology to estimate to define, degree of compacting characterizes the compactness of object, can be described as object bounds girth c and pixel count in object
Ratio between n root mean square；Smoothness characterizes the smooth degree of object bounds, with object bounds girth c and minimum outsourcing square
Ratio between shape girth b is described, and degree of compacting and smoothness are set into weight, calculates the heterogeneous degree of shape facility of object, and root
According to the shape similarity criterion between the change description adjacent object of the heterogeneous degree of the shape facility of object；
The calculation formula of the heterogeneous degree of object shapes feature is：
In formula：wcmpctFor the weight of object degree of compacting, c is object bounds girth, and b is all for the minimum outsourcing rectangle of object
It is long.
It is public according to the calculating of the shape similarity criterion between the change description adjacent object of the heterogeneous degree of the shape facility of object
Formula is：
Step 2.5：In the multidimensional feature space that statistical nature and shape facility are constituted, the statistics that step 2.3 is obtained
The shape similarity criterion that similarity criterion and step 2.4 are obtained is weighted, and then the synthesis calculated between adjacent object is similar
Property criterion, comprehensive statistics feature and the similarity criterion of shape facility as between adjacent object.
The calculation formula of synthesis similarity criterion between adjacent object is：
Δ h=wshapeΔhshape+(1-wshape)Δhstt。
Step 3：The minimum value of similarity criterion between each object and adjacent object is counted, if minimum value is less than or equal to
Scale parameter, then merge two adjacent objects and generate new object, if minimum value is more than scale parameter, nonjoinder, traversal is all
Object, completion is once split, and generates new object layer；
Step 4：Repeat step 2 and step 3 no longer change until object number, obtain final object layer；
Step 5：According to the generation object polygon of step 4, that is, obtain final segmentation result.
With the CIELAB spaces in the polarization characteristic distance replacement simple linear Iterative Clustering based on PauliRGB
Color distance, improves simple linear Iterative Clustering generation super-pixel as initial object as initial segmentation object, both for
The estimation of statistical model parameter provides enough pixels, it also avoid the zigzag phenomenon of partitioning boundary；Based on fractal net work
The thought of evolution algorithmic, combines statistical nature and shape facility to calculate the similarity criterion between object, makes cutting object
More consistent, border is smooth, improves the accuracy of Polarimetric SAR Image segmentation.
The effect of the present invention is further described with reference to analogous diagram.
1. emulation content：
The result of the emulation experiment of the present invention is as shown in Figure 5.Fig. 5 (a) is the Dutch Flevoland that RADARSAT-2 is obtained
Regional C-band Polarimetric SAR Image, range resolution is 4.7m, and azimuth resolution is 4.8m.Fig. 5 (b) is with mesh generation
Method obtains initial object, is then based on the FNEA of K statistical natures segmentation result；Fig. 5 (c) is to be given birth to simple linear iterative algorithm
Into super-pixel as initial object, be then based on the FNEA of K distribution statisticses features segmentation result；Fig. 5 (d) is with simple line
Property iterative algorithm generation super-pixel as initial object, be then based on G0The FNEA of distribution statisticses feature segmentation result.Fig. 5
(c) all it is specific embodiment of the invention with Fig. 5 (d).
2. experimental result and analysis：
From Fig. 5 (b), Fig. 5 (c), Fig. 5 (d) it can be seen that combine causes segmentation with mesh generation method generation initial object
Substantially, segmentation result of the invention is more accurate than more visible to atural object border retention, and uniformity is preferably, side for border zigzag
Boundary is smooth.As can be seen that the segmentation accuracy of Polarimetric SAR Image can be improved using the dividing method of the present invention.
The present invention is using super-pixel as initial segmentation object, and only the parameter estimation of statistical model does not provide sufficient amount of
Pixel, it also avoid the border zigzag phenomenon caused by mesh generation initial object.With the statistics derived based on statistical model
Similitude replaces the spectral similarity in traditional fractal net work evolution algorithmic, according to the similar of comprehensive statistics feature and shape facility
Property criterion complete object merging, so as to realize the Accurate Segmentation of Polarimetric SAR Image.
In the case where not conflicting, the feature in embodiment and embodiment herein-above set forth can be combined with each other.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included in the scope of the protection.
Claims (10)
1. the polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic, it is characterised in that include following step
Suddenly：
Step 1：For Polarimetric SAR Image to be split, generation super-pixel is used as initial object；
Step 2：Calculate each similarity criterion between object and adjacent object in initial object；
Step 3：The minimum value of similarity criterion between each object and adjacent object is counted, if minimum value is less than or equal to yardstick
Parameter, then merge two adjacent objects and generate new object, if minimum value is more than scale parameter, nonjoinder travels through all objects,
Completion is once split, and generates new object layer；
Step 4：Repeat step 2 and step 3 no longer change until object number, obtain final object layer；
Step 5：Object polygon is generated according to step 4, that is, obtains final segmentation result.
2. the polarization SAR image segmentation method according to claim 1 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 1, initial object is used as based on Polarimetric SAR Image generation super-pixel.
3. the polarization SAR image segmentation method according to claim 1 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 1, Pauli is carried out to Polarimetric SAR Image and decomposes generation PauliRGB images；Comprehensive PauliRGB
The polarization characteristic distance and space length of image complete K mean iteratives cluster as final distance metric in subrange,
Generation super-pixel is used as initial object.
4. the polarization SAR image segmentation method according to claim 3 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 1, generates concretely comprising the following steps for super-pixel：
Step 1.1：Determine desired super-pixel size g2, and then step-length g is determined, according to step-length sampling selected seed point, and in 3*
Gradient smallest point of the adjustment seed point to PauliRGB images in 3 subranges；
Step 1.2：Calculated respectively in each central seed point 2g*2g subranges each pixel and correspondence seed point between away from
From；
Step 1.3：Distance between each pixel and correspondence seed point that are obtained according to step 1.2 carries out part K mean iteratives and gathered
Class, until restraining or reaching maximum iteration, that is, generates super-pixel；
Step 1.4：The super-pixel that number of pixels is less than setting numerical value is merged into adjacent and closest super-pixel, produced
To the super-pixel as initial object.
5. the polarization SAR image segmentation method according to claim 4 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 1.1, the PauliRGB values of m-th seed point and locus are labeled as Cm=[Rm Gm Bm xm
ym]T；
In the step 1.2, the calculation formula of distance is between each pixel and correspondence seed point：
<mrow>
<msub>
<mi>d</mi>
<mi>p</mi>
</msub>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>R</mi>
<mi>m</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>G</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>G</mi>
<mi>m</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>B</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>B</mi>
<mi>m</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
<mrow>
<msub>
<mi>d</mi>
<mi>s</mi>
</msub>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>m</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>m</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
<mrow>
<mi>D</mi>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>p</mi>
</msub>
<mrow>
<mi>max</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>d</mi>
<mi>p</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>s</mi>
</msub>
<mi>g</mi>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
In formula：D represents the distance between pixel and correspondence seed point, dpRepresent the polarization characteristic distance decomposed based on Pauli, dsFor
Space length, max (dp) for the maximum of the polarization characteristic distances decomposed of Pauli in the cluster, g is space length in cluster
Maximum, with max (dp) and g respectively to polarization characteristic apart from dpWith space length dsIt is standardized.
6. the polarization SAR image segmentation method according to claim 1 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 2, the similarity criterion between each object and adjacent object is to be based on fractal net work evolution algorithmic
Calculating is obtained, and the similarity criterion between each object and adjacent object combines statistical nature and shape facility.
7. the polarization SAR image segmentation method according to claim 1 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 2, calculates each statistics similarity between object and adjacent object in initial object, build phase
Adjacency pair is as a comprehensive statistics feature and the similarity criterion of shape facility.
8. the polarization SAR image segmentation method according to claim 7 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 2, builds the specific side of comprehensive statistics feature and the similarity criterion of shape facility between adjacent object
Method is：
Step 2.1：Using statistical model to polarization SAR data modeling, and system is estimated according to the probability density function of statistical model
Count the form parameter of model；
Step 2.2：The heterogeneous degree of statistical property of object is calculated using likelihood function；
Step 2.3：Calculate the statistics similarity criterion between adjacent object；
Step 2.4：Will degree of compacting and smoothness setting weight, the heterogeneous degree of shape facility of object is calculated, and according to the shape of object
Shape similarity criterion between the change description adjacent object of the heterogeneous degree of shape feature；
Step 2.5：In the multidimensional feature space that statistical nature and shape facility are constituted, the statistics that step 2.3 is obtained is similar
Property the obtained shape similarity criterion of criterion and step 2.4 be weighted, and then it is accurate to calculate the synthesis similitude between adjacent object
Then, comprehensive statistics feature and the similarity criterion of shape facility as between adjacent object.
9. the polarization SAR image segmentation method according to claim 8 based on super-pixel and fractal net work evolution algorithmic, its
It is characterised by, in the step 2.1, statistical model is G0Distributed model, the probability density function of statistical model is：
<mrow>
<msub>
<mi>p</mi>
<mi>T</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msup>
<mi>L</mi>
<mrow>
<mi>L</mi>
<mi>d</mi>
</mrow>
</msup>
<mo>|</mo>
<mi>T</mi>
<msup>
<mo>|</mo>
<mrow>
<mi>L</mi>
<mo>-</mo>
<mi>d</mi>
</mrow>
</msup>
<mi>&Gamma;</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>L</mi>
<mi>d</mi>
<mo>-</mo>
<mi>&alpha;</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>&Gamma;</mi>
<mi>d</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>L</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mi>&Sigma;</mi>
<msup>
<mo>|</mo>
<mi>L</mi>
</msup>
<mi>&Gamma;</mi>
<mrow>
<mo>(</mo>
<mrow>
<mo>-</mo>
<mi>&alpha;</mi>
</mrow>
<mo>)</mo>
</mrow>
<mi>&Gamma;</mi>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<mo>-</mo>
<mi>&alpha;</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
<mi>&alpha;</mi>
</msup>
</mrow>
</mfrac>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<mi>L</mi>
<mi>t</mi>
<mi>r</mi>
<mrow>
<mo>(</mo>
<mrow>
<msup>
<mi>&Sigma;</mi>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mi>T</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>&alpha;</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
<mrow>
<mi>&alpha;</mi>
<mo>-</mo>
<mi>L</mi>
<mi>d</mi>
</mrow>
</msup>
<mo>,</mo>
<mi>&alpha;</mi>
<mo><</mo>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mrow>
<msub>
<mi>&Gamma;</mi>
<mi>d</mi>
</msub>
<mo>(</mo>
<mi>L</mi>
<mo>)</mo>
<mo>=</mo>
<msup>
<mi>&pi;</mi>
<mrow>
<mi>d</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>d</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msup>
<munderover>
<mi>&Pi;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>d</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<mi>&Gamma;</mi>
<mrow>
<mo>(</mo>
<mi>L</mi>
<mo>-</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
In formula：Σ is the desired value of coherence matrix, its estimateIt can be estimated by the average value of sample：∑=
E [T], L are that, regarding number, α is form parameter, and Γ () is Γ-function, tr () and | | the mark and ranks of matrix are sought in expression respectively
Formula, d is coherence matrix T dimension, the d=3 under the conditions of reciprocal theorem is met；
According to the probability density function of statistical model, and based on tr (∑s-1T second order moment characteristics estimation statistical model probability density)
The calculation formula of form parameter in function is：
<mrow>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mo>=</mo>
<mfrac>
<mrow>
<mn>2</mn>
<mi>L</mi>
<mi>V</mi>
<mi>a</mi>
<mi>r</mi>
<mrow>
<mo>{</mo>
<mi>M</mi>
<mo>}</mo>
</mrow>
<mo>+</mo>
<mi>d</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>L</mi>
<mi>d</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>d</mi>
<mo>-</mo>
<mi>L</mi>
<mi>V</mi>
<mi>a</mi>
<mi>r</mi>
<mrow>
<mo>{</mo>
<mi>M</mi>
<mo>}</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
In formula：M=tr (∑s-1T)。
10. the polarization SAR image segmentation method according to claim 9 based on super-pixel and fractal net work evolution algorithmic,
Characterized in that, in the step 2.2, the heterogeneous degree calculation formula of statistical property of object is：
<mrow>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>t</mi>
<mi>t</mi>
</mrow>
<mi>S</mi>
</msubsup>
<mo>&cong;</mo>
<mo>-</mo>
<mi>n</mi>
<mi>L</mi>
<mi>ln</mi>
<mo>|</mo>
<mover>
<mi>&Sigma;</mi>
<mo>^</mo>
</mover>
<mo>|</mo>
<mo>-</mo>
<mi>n</mi>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mi>l</mi>
<mi>n</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>n</mi>
<mi>ln</mi>
<mo>&lsqb;</mo>
<mfrac>
<mrow>
<mi>&Gamma;</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&Gamma;</mi>
<mrow>
<mo>(</mo>
<mi>L</mi>
<mi>d</mi>
<mo>-</mo>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>&rsqb;</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>L</mi>
<mi>d</mi>
<mo>-</mo>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mo>)</mo>
</mrow>
<msub>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>&Element;</mo>
<mi>S</mi>
</mrow>
</msub>
<mi>l</mi>
<mi>n</mi>
<mrow>
<mo>(</mo>
<mi>L</mi>
<mi>t</mi>
<mi>r</mi>
<mo>(</mo>
<mrow>
<msup>
<mover>
<mi>&Sigma;</mi>
<mo>^</mo>
</mover>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<msub>
<mi>T</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>-</mo>
<mover>
<mi>&alpha;</mi>
<mo>^</mo>
</mover>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula：N is the number of pixels of object；
In the step 2.3, the calculation formula of the statistics similarity criterion between adjacent object is：
<mrow>
<msub>
<mi>&Delta;h</mi>
<mrow>
<mi>s</mi>
<mi>t</mi>
<mi>t</mi>
</mrow>
</msub>
<mo>=</mo>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>t</mi>
<mi>t</mi>
</mrow>
<mi>i</mi>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>t</mi>
<mi>t</mi>
</mrow>
<mi>j</mi>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>t</mi>
<mi>t</mi>
</mrow>
<mrow>
<mi>i</mi>
<mo>&cup;</mo>
<mi>j</mi>
</mrow>
</msubsup>
</mrow>
In formula：I and j represent object i and the object j adjacent with object i respectively；
In the step 2.4, the calculation formula of the heterogeneous degree of object shapes feature is：
<mrow>
<msub>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>h</mi>
<mi>a</mi>
<mi>p</mi>
<mi>e</mi>
</mrow>
</msub>
<mo>=</mo>
<msub>
<mi>w</mi>
<mrow>
<mi>c</mi>
<mi>m</mi>
<mi>p</mi>
<mi>c</mi>
<mi>t</mi>
</mrow>
</msub>
<mfrac>
<mi>c</mi>
<msqrt>
<mi>n</mi>
</msqrt>
</mfrac>
<mo>+</mo>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>-</mo>
<msub>
<mi>w</mi>
<mrow>
<mi>c</mi>
<mi>m</mi>
<mi>p</mi>
<mi>c</mi>
<mi>t</mi>
</mrow>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mfrac>
<mi>c</mi>
<mi>b</mi>
</mfrac>
</mrow>
In formula：WcmpctFor the weight of object degree of compacting, c is object bounds girth, and b is the minimum outsourcing rectangular perimeter of object；
It is according to the calculation formula of the shape similarity criterion between the change description adjacent object of the heterogeneous degree of the shape facility of object：
<mrow>
<msub>
<mi>&Delta;h</mi>
<mrow>
<mi>s</mi>
<mi>h</mi>
<mi>a</mi>
<mi>p</mi>
<mi>e</mi>
</mrow>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<msub>
<mi>n</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>h</mi>
<mi>a</mi>
<mi>p</mi>
<mi>e</mi>
</mrow>
<mrow>
<mi>i</mi>
<mo>&cup;</mo>
<mi>j</mi>
</mrow>
</msubsup>
<mo>-</mo>
<mo>(</mo>
<mrow>
<msub>
<mi>n</mi>
<mi>i</mi>
</msub>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>h</mi>
<mi>a</mi>
<mi>p</mi>
<mi>e</mi>
</mrow>
<mi>i</mi>
</msubsup>
<mo>+</mo>
<msub>
<mi>n</mi>
<mi>j</mi>
</msub>
<msubsup>
<mi>h</mi>
<mrow>
<mi>s</mi>
<mi>h</mi>
<mi>a</mi>
<mi>p</mi>
<mi>e</mi>
</mrow>
<mi>j</mi>
</msubsup>
</mrow>
<mo>)</mo>
<mo>;</mo>
</mrow>
In the step 2.5, the calculation formula of the synthesis similarity criterion between adjacent object is：
Δ h=wshapeΔhshape+(1-wshape)Δhstt。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710411372.7A CN107180434B (en) | 2017-05-23 | 2017-05-23 | Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710411372.7A CN107180434B (en) | 2017-05-23 | 2017-05-23 | Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107180434A true CN107180434A (en) | 2017-09-19 |
| CN107180434B CN107180434B (en) | 2019-10-18 |
Family
ID=59836012
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710411372.7A Active CN107180434B (en) | 2017-05-23 | 2017-05-23 | Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107180434B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090913A (en) * | 2017-12-12 | 2018-05-29 | 河南大学 | A kind of image, semantic dividing method based on object level Gauss-Markov random fields |
| CN108985292A (en) * | 2018-05-23 | 2018-12-11 | 中国地质大学（武汉） | A kind of SAR image CFAR object detection method and system based on multi-scale division |
| CN109544567A (en) * | 2018-11-30 | 2019-03-29 | 中国人民解放军国防科技大学 | Polarized synthetic aperture radar image segmentation method based on two-stage region merging strategy |
| CN110047079A (en) * | 2019-04-26 | 2019-07-23 | 重庆交通大学 | A kind of optimum segmentation scale selection method based on objects similarity |
| CN110827290A (en) * | 2019-10-16 | 2020-02-21 | 中国矿业大学 | Polarized SAR image superpixel segmentation method based on watershed |
| CN111008981A (en) * | 2019-12-26 | 2020-04-14 | 中国人民解放军国防科技大学 | Method, system, device and computer readable medium for segmenting polarimetric synthetic aperture radar image |
| CN116206203A (en) * | 2023-03-08 | 2023-06-02 | 中国石油大学(华东) | Oil spill detection method based on SAR and Dual-EndNet |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103377477A (en) * | 2013-06-27 | 2013-10-30 | 中国科学院遥感与数字地球研究所 | Method and system for multilevel division of high-resolution remote sensing image |
| CN105335975A (en) * | 2015-10-22 | 2016-02-17 | 西安电子科技大学 | Polarized SAR image segmentation method based on low rank decomposition and histogram statistics |
| CN105608692A (en) * | 2015-12-17 | 2016-05-25 | 西安电子科技大学 | PolSAR image segmentation method based on deconvolution network and sparse classification |
-
2017
- 2017-05-23 CN CN201710411372.7A patent/CN107180434B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103377477A (en) * | 2013-06-27 | 2013-10-30 | 中国科学院遥感与数字地球研究所 | Method and system for multilevel division of high-resolution remote sensing image |
| CN105335975A (en) * | 2015-10-22 | 2016-02-17 | 西安电子科技大学 | Polarized SAR image segmentation method based on low rank decomposition and histogram statistics |
| CN105608692A (en) * | 2015-12-17 | 2016-05-25 | 西安电子科技大学 | PolSAR image segmentation method based on deconvolution network and sparse classification |
Non-Patent Citations (2)
| Title |
|---|
| BIN LIU等: ""Superpixel-Based Classification With an Adaptive Number of Classes for Polarimetric SAR Images"", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 》 * |
| 刘修国等: ""选择利用Wishart和K统计描述的极化SAR图像分割"", 《电子科技大学学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090913A (en) * | 2017-12-12 | 2018-05-29 | 河南大学 | A kind of image, semantic dividing method based on object level Gauss-Markov random fields |
| CN108985292A (en) * | 2018-05-23 | 2018-12-11 | 中国地质大学（武汉） | A kind of SAR image CFAR object detection method and system based on multi-scale division |
| CN109544567A (en) * | 2018-11-30 | 2019-03-29 | 中国人民解放军国防科技大学 | Polarized synthetic aperture radar image segmentation method based on two-stage region merging strategy |
| CN110047079A (en) * | 2019-04-26 | 2019-07-23 | 重庆交通大学 | A kind of optimum segmentation scale selection method based on objects similarity |
| CN110827290A (en) * | 2019-10-16 | 2020-02-21 | 中国矿业大学 | Polarized SAR image superpixel segmentation method based on watershed |
| CN111008981A (en) * | 2019-12-26 | 2020-04-14 | 中国人民解放军国防科技大学 | Method, system, device and computer readable medium for segmenting polarimetric synthetic aperture radar image |
| CN116206203A (en) * | 2023-03-08 | 2023-06-02 | 中国石油大学(华东) | Oil spill detection method based on SAR and Dual-EndNet |
| CN116206203B (en) * | 2023-03-08 | 2023-08-18 | 中国石油大学(华东) | Oil spill detection method based on SAR and Dual-EndNet |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107180434B (en) | 2019-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107180434A (en) | Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic | |
| CN101551905B (en) | Method for segmenting multi-dimensional texture image on basis of fuzzy C-means clustering and spatial information | |
| CN104049245B (en) | Urban building change detection method based on LiDAR point cloud spatial difference analysis | |
| CN113781667B (en) | Three-dimensional structure simplified reconstruction method and device, computer equipment and storage medium | |
| CN106709948A (en) | Quick binocular stereo matching method based on superpixel segmentation | |
| CN111275724B (en) | Airborne point cloud roof plane segmentation method based on octree and boundary optimization | |
| CN105608692B (en) | Polarization SAR image segmentation method based on deconvolution network and sparse classification | |
| CN104794730B (en) | SAR image segmentation method based on super-pixel | |
| CN109410342A (en) | A kind of point cloud compressing method retaining boundary point | |
| CN104050628B (en) | Image processing method and image processing device | |
| CN105138970A (en) | Spatial information-based polarization SAR image classification method | |
| CN108154104A (en) | A kind of estimation method of human posture based on depth image super-pixel union feature | |
| CN106097252B (en) | High spectrum image superpixel segmentation method based on figure Graph model | |
| CN103745472B (en) | SAR image segmentation method based on condition triple Markov field | |
| CN111667574B (en) | Method for automatically reconstructing regular facade three-dimensional model of building from oblique photography model | |
| CN110163213A (en) | Remote sensing image segmentation method based on disparity map and multiple dimensioned depth network model | |
| CN106780508A (en) | A kind of GrabCut texture image segmenting methods based on Gabor transformation | |
| CN106228594A (en) | Typhoon model cloud cartoon display method based on surface subdivision | |
| CN110807781A (en) | Point cloud simplification method capable of retaining details and boundary features | |
| CN107564049A (en) | Faceform's method for reconstructing, device and storage medium, computer equipment | |
| CN108734228A (en) | The polarimetric SAR image random forest classification method of comprehensive multiple features | |
| CN102073867A (en) | Sorting method and device for remote sensing images | |
| Koppel et al. | Sentinel-1 for urban area monitoring—Analysing local-area statistics and interferometric coherence methods for buildings' detection | |
| CN116189147A (en) | YOLO-based three-dimensional point cloud low-power-consumption rapid target detection method | |
| CN108846845A (en) | SAR image segmentation method based on thumbnail and hierarchical fuzzy cluster |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |